摘要
交流电动机转子断条故障诊断的问题可以归结为模式分类的问题,支持向量机是近几年来涌现的一种新型的针对小样本集合具有较好分类效果的方法。根据断条故障在定子电流谱中产生相应特征分量,本文利用窗提取的方法对定子电流谱进行转换,在不损失故障特征信息的前提下构造低维数特征空间,并在其中用SVM进行分类。同时针对实际应用中转差率s由于负载的变化而使故障特征频率位置发生变化的问题,分别在恒定负载和变负载条件下进行试验。断条情况为无断条、一根断条、连续两根断条三类,结果表明在恒定负载和变负载条件下都有较高的诊断正确率。同时因为整个诊断过程可以实现自动化,所以可实现在线诊断。
The diagnosis of AC motor broken rotor bar fault falls into the category of pattern classification in some respects. Support vector machine （SVM） has good performance for classification over small sample set. This paper presents a SVM-based method of fault diagnosis of broken rotor bar in induction motor. As broken rotor bar causes frequency components in stator current power spectra, feature data sets are constructed from power spectra of stator current with a windowing method which reduces the dimension of feature sets efficiently and keeps the fault feature information. Because fault feature frequency varies as load changed in practice, experiments are taken with stable load and with unstable load respectively. Motor rotor conditions tested are health, one broken bar, two neighboring broken bar. Two SVM-based classifiers are constructed; one classifier detects health, one or two broken rotor bar and the second classifier detects fault or health. Experiments show the results are promising for each condition, which means SVM are practical for use to detect broken rotor bar fault. And because the procedures can be finished automatically, it is suitable for online diagnosis system.
出处
《电工技术学报》
EI
CSCD
北大核心
2006年第8期48-52,共5页
Transactions of China Electrotechnical Society
基金
国家自然科学基金资助项目(50307011)。
